Survey of Image Based Graph Neural Networks

Another example is making inferences about facial attributes and Deep learning, particularly convolutional neural networks have in identify by representing facial landmarks as a graph [33]. The literature the recent past revolutionized many machine learning tasks. Examples on application of GNNs on images can be broadly classified in to include image classification, video processing, speech recognition, three groups (a) pixel-based graphs, (b) superpixel-based graphs and and natural language processing. These applications are usually (c) object-based graphs - sample illustrations of these three methods characterized by data drawn from the Euclidean space. Recently, are shown in Figure 1. In addition to providing a comprehensive review many studies on extending deep learning approaches for graph data of graph techniques for images' superpixels, this paper paper have emerged [1-22]. The motivation for these studies stems from also makes notable contribution by introducing new taxonomy based the emergence of applications in which data is drawn from noneuclidean on how graph represents an image as summarized in Table 1.